billbr@xstor.UUCP (Bill Brothers) (06/04/90)
In article <6735@ccncsu.ColoState.EDU> roh@handel.UUCP writes: >Does anyone know what the real differences are between the cheap (<$100) and >expensive (~$300) SCSI adaptors? Well, I must generalize here, so bear with me. Expensive controllers most often have "intelligence". In other words there is an on-board processor. (That makes the the board considerably more expensive to design and manufacture.) This processor relieves the CPU (MPU for technoids :-) from the overhead of watching all the the states and conditions happening on the SCSI bus. This is most benificial in an OS that is multi-threaded. Under DOS, there is so much time spent idle, increased overhead is usually not visible. Inexpensive boards usually contain "no intelligence". This means that all state transitions, ack/nack, and other garbage must be done by the device driver/host cpu. >Are the expensive ones that much better? Generally yes. >If so, in what ways? The board manufacturer is generally "up to speed" on what is happening in the SCSI world. Having a "high level" hardware interface helps to lengthen the life of the device driver in the ever-changing hardware world. >Since most scsi drives have 8bit interface, what good >is it to have a 16bit host adaptor. Basically, (no flames please for this over-simplification) SCSI can run faster than the AT ISA bus. When operating SCSI in burst mode, many AT's just can't handle the tranfer rate without saturating the bus. This is not true for all ATs but for most. There also comes into play something called first-party DMA. This means that the host adapter takes over the bus and shoves data into memory (Usually at a much higher rate of speed that the motherboard DMA). >Also does the adaptors have anything to >do with the interleave (1:1 and 2:1 adaptors exist?). Generally not. They can under rare conditions, but mostly the controller on the drive takes care of it. SCSI is a very different beast in this respect. IF implemented properly, several drives can be operated simul- taneously under SCSI. >Yes, these are lots of questions and I would appreciate some response. >Thanks, >Lucas I hope this has been useful. Please feel free to mail me if you have other questions. ---- Bill Brothers Product Engineering Mgr. Storage Dimensions, Inc. Voice (408) 379-0300 uunet!xstor!billbr billbr@xstor.UUCP